Sentiment & Resolution
Understand how visitors feel and whether their questions get answered.
Two of the most important cards on the Analytics tab tell you how visitors feel about their conversations and whether they are getting the answers they need.
User Sentiment
The User Sentiment card shows how visitors felt during their conversations with your companion. Every conversation is automatically analyzed and categorized as:
- Positive -- The visitor seemed happy, satisfied, or engaged. This is what you want to see the most of.
- Neutral -- The visitor was neither particularly happy nor unhappy. Typical for straightforward, factual exchanges.
- Negative -- The visitor seemed frustrated, confused, or dissatisfied.
The card displays a donut chart with the breakdown and shows a satisfaction rate percentage at the center. This rate combines positive and neutral conversations as a share of the total -- the idea being that neutral is perfectly fine, while negative is the signal to watch.
What to do with this
If you see a high percentage of negative conversations, it usually means your companion is struggling with certain types of questions. Check the Top Keywords card to see which topics come up most, and consider adding more knowledge or adjusting behaviour rules to handle those areas better.
You can also check the User Friction signals to see if users are experiencing repeated questions, clarification loops, or task abandonment — these often correlate with negative sentiment.
Resolution Rate
The Resolution Rate card tells you whether visitors' questions are actually getting answered. Each conversation is tagged as:
- Resolved -- The visitor's question was fully answered.
- Partially resolved -- The companion gave a partial answer but could not fully address the question.
- Unresolved -- The companion could not answer the question at all.
The card shows a donut chart and a resolution rate percentage. This rate gives full credit for resolved conversations and half credit for partially resolved ones.
Failure reasons
When a conversation is unresolved or partially resolved, Aura identifies why it failed. See the Failure Analysis card (Row 2) for a full breakdown by root cause and severity.
Common failure reasons include:
- Knowledge gap -- Your companion did not have the information needed. Upload more resources to fix this.
- Out of scope -- The question was outside your companion's domain. This is normal and expected.
- Unclear question -- The visitor's question was too vague for the companion to understand.
- Hallucination -- The companion generated an incorrect response. Rare but important to watch.
- Misunderstood intent -- The companion misinterpreted what the user wanted.
- Technical error -- Something went wrong on the technical side.
Response Quality
The Response Quality card (Row 1) scores your companion across five dimensions on a 1--5 scale:
- Relevance -- Was the response about what the user asked?
- Correctness -- Was the information accurate?
- Completeness -- Was the answer thorough enough?
- Clarity -- Was the response easy to understand?
- Tone Fit -- Did the tone match the situation?
The scores are displayed as a radar chart with an overall letter grade (A through E). Focus improvement efforts on your lowest-scoring dimension.
Last updated on